Predicting tags for StackOverflow questions through a multi-class multi-tag classifier system

Abstract

. This work approaches the contest “Stack Overflow Tag Prediction” published on Kaggle website. The challenge is to identify suitable tags for question on StackOverflow. This can be termed as a multi-class multi-tag classification problem, as a question can come under different topics and also have several tags. The solution proposed is three – way classifier system where the question is assumed to have a maximum of three tags. To understand the feasibility of this method we used 3 models of classification namely - SVM, Naïve Bayes and Logistic Regression and compared the results using certain score metrics.